摘要
当前的程序设计都是人工设计执行流程,这种方法具有被动性、机械性、缺乏灵活性等缺点。提出一种基于强化学习的程序设计机制,并实现了相应的算法。根据环境情况和问题要求让计算机自主选择执行流程,通过学习使结果达到最优,同时能实现分层调用。采用这种方法,程序执行可以自主决策,较好地实现了自适应,减少了对设计者的依赖。结果显示,这种方法能取得较好的运行效率。
The current computer programming is designed artificially,which has the weakness of passivity,rigidity and lack of flexibility.This paper proposes a method based on reinforcement learning mechanism,and realizes the corresponding algorithm.According to the environment and requirements,the agent can choose executive process independently and arrive at the optimal result by learning,realize the layered calls.Using this method,the executing program is decision-making,has a way to realize the adaption,and reduces the dependence on designer.The result shows that the method can achieve satisfactory execution efficiency.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第36期80-82,126,共4页
Computer Engineering and Applications
基金
山东省自然科学基金(No.ZR2009GM009)
陕西省教育厅专项科研计划项目(No.08JK430)
关键词
自适应程序设计
强化学习
Q学习
AGENT
优化算法
adaptive programming
reinforcement learning
Q-learning
agent
optimization algorithm
